Research On The Key Technology Of Intelligent Fault Tolerance In Vision Guidance System Of AGV

The continuous development of logistics automation technology promotes the wide application of automated guided vehicle(AGV) in manufacturing, warehousing, logistics and other industries.In order to adapt to the complex working environment of the industrial field, the AGV needs to have a certain tolerance to various factors that may cause the disturbance of the vision guidance system.On the base of summing up the existing AGV vision guidance technology,this paper thoroughly studies the technology of intelligent fault tolerance for illumination,defaced image,occluded image and jitter image,path feature robust recognition technology and precise positioning technology of AGV.Firstly,for the influence factors of light, this paper proposes an adaptive dimming technology of LED ring array light source, which uses PWM wave to adjust the brightness of the light source, so that the collected image has a constant average brightness.For the LED ring array light source,its light model is established, the optimum installation height of the light source with uniform illumination is determined, and the range of high light area and low illumination area is established.For the case of light source occlusion or partial LED lamps failure,a light source state monitoring method based on image is presented in this paper,which can determine the LED dead region by using frame difference method.Secondly,for the case of the failure of path image segmentation due to the defaced ground path,this paper puts forward a defacement self-remediation method based on morphplogy which effectively solve the interface to the path feature extration.For the case of the crack or occlusion of the guidance path,an image occlusion processing method based on least square fitting metnod is proposed.This method takes the least square method to fit the sttaight guidance path and the arc turning path,which realize the reliable guidance of AGV by using the data of the fit line or fit arc.The camera is easy to shake in the process of AGV operation, an image anti-shake algorithm is proposed to filter out the large jitter of AGV and ensure the smooth operation of AGV.Thirdly,to improve vision-guided AGV's reliable recognition rate of multi-branch path and station flag,we take the whole image as input features, and map the features to high-dimensional space and reduce its dimensionality by KPCA, then use BP Neural Network to recognize the sample matrix after dimensionality reduction.The pattern recognition window and the guidance scanning window are set up in the field of view.The path is simplified to a linear model in the guidance scanning window, and the operation speed of the algorithm is improved.Twice visual positioning method is proposed for the precise positioning of vision-guided AGV which can eliminate the adverse effects of inertia on the precise positioning and achieve millimeter level positioning of AGV.Finally, this paper develops a vision guided AGV based on Mecanum wheel platform, and carries out a large number of experiments on this basis. The experimental results verify the effectiveness of the intelligent fault tolerant vision guidance system constructed in this paper.